Pansharpening Methods Based on ARSIS Concept

Pan-sharpening aims to use image fusion techniques in the remote sensing field in order to synthesis the Multispectral (MS) images to higher resolution using spatial information of the Panchromatic (Pan) image. Up to now, several definitions for the image fusion have been suggested. Wald’s definition (Wald, 1999) is one of these most celebrated definitions used commonly in the remote sensing community which defines image fusion as: ”a formal framework in which are expressed means and tools for the alliance of data originating from different sources. It aims at obtaining information of a greater quality, although the exact definition of ‘greater quality’ will depend on the application”. Many applications such as feature detection, change monitoring, urban analysis, and land cover classification recieve benefits of pan-sharpening. In fact, these applications need both high spectral and spatial resolution concurrently. Due to physical and technological constraints, creating a sensor which can provide high spectral and spatial resolution simultanously is not possible. So, the image fusion algorithms have been received increasingly attention to fuse MS and Pan images and to provide a new image including both spatial charachteristics of Pan and spectral charachteristics of MS images. Usually the pan-sharpening methods are categorized into three main sets (Wald, 2002; Thomas et al., 2008);projection substitution, relative spectral contribution, and methods that belong to the Amelioration de la Resolution Spatiale par Injection de Structures (ARSIS) concept. The Projection–Substitution methods take advantage of a vectorial algorithm. In this kind of methods, all fused images corresponding to different MS images are synthesized simultaneously. These methods consider coincident pixels of MS images as spectral axes. Then, the spectral axes are projected into a new space to reduce the information redundancy. It results the decorrelated components. The structures of MS images, which are mainly related to color, are isolated by one of these components from the rest of the information. Actually these methods assume that the structures contained in this structural component are equivalent to those in the Pan image. Next, this structural component is replaced either partially or wholly with corresponding parts of Pan. Eventually, the inverse projection is performed to obtain the MS images in higher resolution, i.e. the fused images. The most famous methods of this category are those based on principal component analysis (PCA) (Ehlers, 1991; Chavez et al., 1991) and intensity hue saturation (IHS) (Haydn et al., 1982). The Relative Spectral Contribution methods are also based on the linear combination of bands. The basic assumption of these methods is considering the low-resolution Pan as a